Abstract: Latinos receive more aggressive, burdensome end-of-life (EOL) care (eg, ICU stays, resuscitation) and less hospice care than non-Latino whites. The available evidence suggests that the EOL care Latinos receive may be suboptimal and inconsistent with their wishes, and inferior to the EOL care that whites receive. The overarching aim of this study is to identify the most promising targets for interventions designed to enable Latinos to receive: a) high quality EOL care, and b) care consistent with their values and preferences ("treatment goal attainment"). Our preliminary results, and those of others, suggest that there is a critical need for data at institutional, provider, and patient levels so that their relative influence can be discerned. The primary aims ofthe proposed study are to obtain multi-level data and use hierarchical linear modeling (HLM) to estimate patient, provider and institutional effects on Latino-white disparities in EOL care and treatment goal attainment. We hypothesize the primacy of patient and provider over institution effects, which will be significant, but less influential than either patient or provider effects. We will recruit 250 advanced gastrointestinal and thoracic cancer patients (125 Latino, 125 non-Latino white) with a life-expectancy of less than 6 months from five sites across the US. We will also enroll 50 oncology providers overall who each care for at least 5 study participants The patient's medical care received in the last month of life will be documented via medical chart extraction in the postmortem assessment. We anticipate that this study will inform policy makers and institutional leadership of where they should invest for the greatest "bang for the buck" to reduce Latino-white disparities in EOL care. The study team, comprised of Dr. Holly Prigerson, a leading expert in EOL care, and Dr. Jan Mutchler, a nationally recognized gerontologist with strong interests in disparities, is well-poised to undertake this work, in collaboration with a junior investigator at DFCI (Jimenez) and a UMB Associate Professor who wants to increase her research skills and expertise (Rivera). This project will benefit significantiy from support provided by the Training and Survey and Statistical Methods Cores.